> ## Documentation Index
> Fetch the complete documentation index at: https://docs.open-hive.com/llms.txt
> Use this file to discover all available pages before exploring further.

# CRM

> The system of record your agents read from and write to.

The **CRM** holds the people, companies, and deals your agents work on. It isn't a
separate system you sync with — it's the shared state your colonies read from and write
to as they run.

<Frame caption="Setup is a guided conversation with a RevOps queen, not a form.">
  <img src="https://mintcdn.com/open-hive/1xpysC1e-PlVBDLW/images/crm.png?fit=max&auto=format&n=1xpysC1e-PlVBDLW&q=85&s=4a3c2f10171970920c13fe6d5b15a0fc" alt="Setup is a guided conversation with a RevOps queen, not a form." width="1440" height="900" data-path="images/crm.png" />
</Frame>

## Setup

The CRM needs a one-time setup per workspace. Until that's done, the section shows a
setup prompt instead of a grid.

Setup is a guided conversation rather than a form: a RevOps queen walks you through
mapping your sales pipeline, its stages, and your ideal customer. It takes about five
minutes. If you'd rather look around first, **Explore the CRM anyway** opens the section
without committing to setup.

<Note>
  [Inbox](/guides/inbox) stays reachable whether or not the CRM is set up. The two are
  gated separately, so an unconfigured CRM never hides your mail.
</Note>

## The four views

| View              | Holds                                                               |
| ----------------- | ------------------------------------------------------------------- |
| **People**        | Individual contacts.                                                |
| **Organizations** | Companies and accounts.                                             |
| **Opportunities** | Deals in your pipeline.                                             |
| **Activity**      | What happened — touches, replies, and events against those records. |

Each view is schema-driven: its columns come from the server, so what you see is the
business-level record. Raw tables, internal IDs, and metadata blobs never reach the
grid.

## Browsing records

* **Filter** on any column to narrow what's shown.
* **Group** records into a board — grouping Opportunities by stage gives you a pipeline
  view.
* **Scroll** to page through large sets. Records load in pages rather than all at once,
  so a big CRM stays responsive.

## Reading and writing

The grid is **read-only**. You browse, filter, and inspect there, but you don't edit
cells directly.

Writes happen through your agents. Ask a colony to update a record, move a deal, or load
a set of accounts, and it writes through the CRM agent path. This is deliberate: every
change to the system of record arrives with an agent's reasoning attached, rather than
as an untracked manual edit.

## Watching agents build your CRM

When agents are writing, the board refreshes itself — you can watch records appear as a
colony seeds them, instead of reloading to find out whether anything happened. Bursts of
agent writes are batched into a single refresh, so a colony seeding a hundred records
doesn't flicker a hundred times.

## Where records come from

* **[Foundry](/guides/foundry) runs** — the CRM plan artifact declares the stages and
  columns a motion needs, and matched buyers load into the first open stage.
* **Colonies** — sourcing and enrichment work writing accounts and contacts as it
  confirms them.
* **[Inbox](/guides/inbox)** — the mail bridge writes conversations back onto matching
  records, so a reply shows up against the person who sent it.